Wrapped Bitcoin: what moves the reserve in Liquid, WBTC and tBTC

Measured against the only known peg address, L-BTC (Liquid's wrapped bitcoin) has a backing of 85.09% as of September 13, 2026, which rises to 85.76% when adding peg-in deposits accepted since the restart that have not yet been swept to that address, according to our own measurement via Blockstream's public APIs. There are 4,232.39358347 L-BTC in circulation; the address holds 3,601.47207156 BTC, and another 28.39144647 BTC are waiting in 14 pending deposits. The gap—between 602.53 and 630.92 BTC, worth 46.44 to 48.63 million dollars at $77,071.13 per bitcoin—opened on a network that requires 11-of-15 signatures to move a single satoshi, without any keys being compromised. This detail is what makes the case comparable: it shows exactly what is required for reserves to move in each wrapped bitcoin model (BTC represented as a token on another chain). This article compares the six active models—federation with hardware security modules (HSM), single custodian, threshold signature network, distributed custody via multi-party computation, and staking without reserves—traces the 3,996 BTC that left on-chain and the 85% that has already returned, and measures how the rest of the complex performed between September 5 and 13, 2026, in bitcoin units.

How much backing does Liquid Network's L-BTC have on Sep-13-2026?

The L-BTC supply is not read from a contract: it is reconstructed using the accounting published by the asset's own explorer. On September 13, 2026, the Blockstream API for the L-BTC asset shows 18,385.21976870 BTC in cumulative peg-ins (bitcoin delivered to the federation to receive L-BTC), 14,142.79839199 BTC in cumulative peg-outs (the reverse operation), and 10.02779324 BTC burned. The subtraction leaves 4,232.39358347 BTC of L-BTC in circulation.

On the other side, the federation address—bc1qdlld6ant…jwaxxsuhwxxr, identified by the Bitquery investigation and the one that received the returns—has received 23,362.86333008 BTC and spent 19,761.39125852 BTC across 597 transactions; its last spend was on September 6, 2026, at 16:48:45 UTC. Its balance is 3,601.47207156 BTC. Taken as the sole peg address with a balance, the ratio yields a backing of 85.09%.

This direct measurement has a bias that can be observed on-chain since the restart. A Liquid peg-in does not deposit bitcoin directly into that address: it deposits it into an address derived from the federation's script, and the federation later sweeps it into the reserve. Before the halt, sweeping was routine: of 40 deposits claimed on Liquid between September 4 and 6, 2026, all 40 ended up in bc1qdlld…, with a median delay of 19.2 hours from their confirmation on Bitcoin. Since the restart, none have been swept. The 14 peg-ins claimed between September 9 at 21:11:10 UTC and September 13 at 08:27:10 UTC remain in their deposit addresses and total 28.39144647 BTC: bitcoin on-chain, under the peg-in's multi-sig by design, but outside the known address. Adding these, and accounting for a 0.001 BTC peg-out processed on Liquid on September 11 that is still unpaid, the backing is 85.76%. Ten of those deposits—25.32314581 BTC—already existed on September 11, 2026, so the figure provided by the same direct measurement that day (85.15%) carried the same bias: adjusted, it was 85.75%.

The bias grew because peg-ins have resumed. SideSwap, the federation member whose authorization key was used for the September 6 peg-out, reopened peg-ins on September 11, 2026, but keeps peg-outs closed "while the Liquid Federation completes its security review," according to its own notice. Between September 11 and 13, 3.06830066 BTC entered through this route, and none have reached the known address.

Both figures depend on an assumption that should be made explicit: that the federation does not maintain a balance in another reserve address. Blockstream does not publish an inventory of addresses, and its help center describes the destination of peg-ins as "an address secured by the federation's 11-of-15 multi-sig," in the singular. The adjusted figure is also a floor, as it does not include deposits sent to Liquid but not yet claimed. If more balances were to appear, the percentage would rise, but the comparative argument holds regardless, as it depends not on the size of the gap but on how it was opened.

The deficit allows for four different measurements, and they are not averaged.

The range goes from 596.02 to 630.92 BTC, with the unadjusted method being the outlier: once pending deposits are added, the other three are within 6.52 BTC of each other. As of September 13, 2026, there is between 0.85 and 0.86 BTC of backing for every L-BTC, measured against the only known address and its pending deposits, and Blockstream CEO Adam Back has stated that the 1:1 parity will be covered. SideSwap published its own figure on September 10—4,205 L-BTC in circulation against 3,597 BTC in reserve—which does not match the explorer's accounting and is not integrated with it.

What is required to move the reserve in Liquid, WBTC, cbBTC, tBTC, or Babylon?

The useful axis for comparing wrappers is not brand or yield: it is the list of conditions that must be met for the stored bitcoin to leave its location. These are the six active models, with supply measured on September 13, 2026.

The figures for WBTC, cbBTC, tBTC, and FBTC represent the totalSupply() of their Ethereum contracts, read at block 25,969,446 via public RPC (the standard access point for querying a node); they do not include portions issued on other chains, which for WBTC adds approximately 380 more BTC according to aggregators. The Liquid figure comes from the asset explorer's accounting; the Babylon figure is the series calculated by DefiLlama based on the locking scripts users create on the Bitcoin chain itself. FBTC and tBTC reappear further down with different figures—8,584.31 and 4,394.30 BTC—because the basket measures their aggregate supply across all chains where they circulate, beyond Ethereum.

The question of who signs your exit organizes this comparison just as it did for the LayerZero bridges in August 2026. Ranked by the number of required signatures, Liquid holds the second spot for strictness, behind tBTC and far ahead of cbBTC, where an internal decision suffices. And yet, Liquid is the one that lost 95% of its reserve in a single afternoon. Quorum size did not dictate the risk.

Why didn't eleven-of-fifteen signatures stop the outflow of 3,996 BTC from the Liquid federation?

Because on September 6, 2026, all fifteen Liquid officials were running the same copy of Elements, the software that powers the sidechain—the parallel chain anchored to Bitcoin—and each signs when their copy validates the state of the chain. Eleven-of-fifteen signatures are not eleven independent opinions on whether a transaction is legitimate. If the software accepts a proof as valid when it is not, all fifteen accept the same thing simultaneously, and the threshold is reached without anyone bypassing procedure.

Elements version 23.3.4 includes pull request 1600, titled "sigcache: harden range proof cache keys and add -norangeproofcache option," created on September 8, 2026, at 17:19:54 UTC and merged at 19:06:54 UTC the same day. Its description identifies the flaw: the cache storing already validated range proofs constructed its key by concatenating raw fields, such that "distinct argument tuples with byte-identical concatenations" collided in the same entry. Range proofs underpin Liquid's confidential transactions: they prove that a hidden amount is positive and consistent without revealing it. A key collision in that cache allows a false proof to be accepted as valid because a different one already was.

Translated to the balance sheet: it was possible to claim bitcoin that never entered. The Liquid Federation incident report, published on September 8, 2026, at 19:10 UTC, describes the path: the peg-out was processed via the peg-out authorization key (PAK) of SideSwap, a federation member, and both the SideSwap node and the distributed officials accepted the presented L-BTC as valid. The explorer's accounting matches this: cumulative peg-in and cumulative peg-out grow by the same amount, the calculated supply does not move, and the reserve drops.

What has been confirmed by Back is what did not happen: no official or PAK keys were compromised. The failure lay in the validation that the multi-sig accepts as good.

The version running when the incident occurred had been public for months. Elements 23.3.3 was released on April 13, 2026, at 16:21:46 UTC, and Elements 23.3.4, which brings the range proof cache hardening, was published on September 9, 2026, at 03:14:44 UTC. Between the two, 148 days and 11 hours passed; the previous jump, from 23.3.2 to 23.3.3, had been 55 days. The press places the patch at 13:30 UTC that day, which is the time of the announcement, not the release tag on GitHub.

In the repository history, there is a coincidence anyone can verify. Pull request 1599, "Cherry picks from 23.x into 23.3.x," was merged on September 6, 2026, at 17:21:03 UTC—three hours after the bitcoin outflow—and carries a commit dated August 3, 2026, at 10:53:50 UTC, "fix: range proof cache bind to asset and scriptpubkey," which modifies src/script/sigcache.cpp with three lines added and three deleted. This is the same file that PR 1600 rewrites five weeks later with 79 lines added and 24 deleted. The coincidence goes beyond the subject: it is the file itself. This does not authorize saying the fix already existed 34 days earlier: the August commit adds the asset and scriptPubKey to a key that was still being constructed by raw concatenation, and PR 1600 argues the problem lies in that concatenation. It is consistent with both having corrected it earlier or having expanded the attack surface, and no public source dictates which of the two occurred.

How much bitcoin left Liquid on Sep-6-2026 and how much has returned?

The press has valued it at approximately 320 million dollars and called it a hack; on-chain, 3,996.01834922 BTC left and 3,404.00018504 have been returned as of September 13, 2026. The sequence is reconstructed using Blockstream's Bitcoin and Liquid APIs, GitHub history, the federation's incident report, and the decoding of OP_RETURN messages—the field in a Bitcoin transaction that allows recording arbitrary data—written to the peg address. There is also a record of this conversation in a gist by Sjors Provoost, automatically generated and not verified in detail by its author; the quotes are cross-referenced against our own transaction decoding.

The chain served by the explorer today is no longer the one that existed on September 6, and that changes how any time measurement is read. Between block 4,050,335 (Sep-6 13:52:10 UTC) and the one currently occupying height 4,050,336 (Sep-9 21:05:10 UTC), there is a 79-hour and 13-minute gap, which represents the halt plus the window of discarded blocks. Liquid continued producing blocks after 13:52:10: the original 4,050,336, dated 13:53:10, contained the peg-out, and 4,050,349, from 14:06:10, processed the request for 3,996.01834922 BTC via the SideSwap PAK. The chain stopped at height 4,051,232 on September 7 at 04:49 UTC—the same resource used by THORChain in its 13-hour halt in May 2026—and those 897 heights disappeared in the restart—our own calculation based on the difference between the halt height and the first discarded block. The actual halt lasted 64 hours and 16 minutes, from Sep-7 at 04:49 to Sep-9 at 21:05:10 UTC. As of September 13, 2026, the explorer continues to serve those heights with post-restart dates: there has been no second reorganization.

The time of that block has two versions and they are not averaged: the Liquid Federation incident report places it at 15:53:10 UTC, Bitquery and Defiprime date it at 13:53:10 UTC, and the current canonical chain does not preserve the block that would decide. The two-hour difference allows for explanations, none of which have been published by the federation.

Two more readings emerge from the sequence and do not depend on which version of the chain is viewed. First: in addition to the 3,400 BTC, on September 9, 4.00000000 BTC entered the peg address from three addresses with no on-chain link to those that received the payment, plus 18,504 satoshis (0.00018504 BTC) distributed in 23 microtransactions with OP_RETURN messages. The total deposited between September 7 and 11 is 3,404.00018504 BTC, and until September 13, only dust has entered. Second: the net effect on the reserve from the 14:28:56 transaction was −4,002.66609035 BTC, slightly more than the major outflow of 3,996.01834922, because that transaction has thirteen outputs and returns part to the federation itself as change.

The address holding the funds—bc1ql4mfu6…, the second in the forwarding chain—holds 598.50426749 BTC as of September 13, 2026, and has not sent bitcoin to third parties again: its last outflow is a self-consolidation on September 9 at 11:45:49 UTC accompanying the 10% request message. It was created at 14:01:57 UTC on September 6, 27 minutes before the drain, and has reconsolidated the same funds dozens of times, such that its 28,779.00 BTC received is the same money counted many times; of its 1,430 transactions, the 1,133 recorded since September 11 are all incoming dust and messages. The request for a 10% reward would equal 399.60 BTC based on the 3,996.02 that left.

Was there contagion to WBTC, Babylon, or tBTC following the Liquid event?

The answer is unclear in dollars but clear in bitcoin units. These are the nine largest BTC wrappers and derivatives listed on aggregators, measured using the token series published by DefiLlama per protocol, excluding debt and internal staking series.

The last available data point for Lombard LBTC is from September 12, 2026, at 21:22 UTC; the rest of the series close on September 13 between 13:48 and 14:16 UTC.

The basket lost 0.175% of its units between September 5 and 13, 2026. Removing exSat, which accounts for 86% of the drop, the rest moved by 0.025%: 47.75 BTC out of 193,608. In dollars, the same basket went from 15,606 to 15,080 million, a −3.37% change, while the price of bitcoin fell from $79,616.73 to $77,071.13, a −3.20% change. Almost the entire dollar drop is due to the price of bitcoin.

The exSat movement cannot be blamed on Liquid: its daily series dropped for three consecutive days—2,392.44 BTC on September 7; 2,365.15 on the 8th; 2,342.14 on the 9th—rebounded to 2,382.87 on the 10th, and fell again to 2,106.27 on September 13, a week after the event. And Babylon, the model that has no reserve to move because the bitcoin does not leave the user's custody, added 39.99 BTC between September 5 and 13, 2026.

Why doesn't Liquid appear in DefiLlama's BTCFi rankings?

Because it isn't there. As of September 13, 2026, DefiLlama's protocol list has 8,243 entries and none declare "Liquid" among their chains; the chain list has 467 and the only one containing that word is "Hyperliquid L1," which is unrelated. The 4,232.39358347 BTC of L-BTC are worth 326.20 million dollars at $77,071.13 per bitcoin on September 13, 2026, which is 1.94 times the total value locked that DefiLlama counts for Rootstock (77.88 million), Stacks (79.73 million), and BOB (10.28 million)—167.89 million combined. These are different bases and it must be stated: all backed supply on one side, value locked in decentralized finance applications on the other. And of those three, BOB is a hybrid L2 on Ethereum, so it doesn't even fall into Liquid's category.

The practical consequence affects the comparison method: any "BTCFi size" figure—decentralized finance built on bitcoin—taken from an aggregator excludes the largest bitcoin sidechain by wrapped value, which is also the only one with a reserve failure traced on-chain in 2026. A comparison built on that table works with an incomplete sample.

What can a wrapped bitcoin holder check before choosing a model?

Four checks that can be done from the outside, without privileged access and using the same APIs used here:

1. Where the reserve is and if it can be measured. If the issuer publishes addresses, the balance is a simple query. If they do not—as in the case of cbBTC—the guarantee is contractual and accounting-based, not observable on-chain. And even if they are published, one must check that the measurement includes deposits in transit: in Liquid, between September 9 and 13, 2026, 28.39 BTC of peg-ins remained outside the known address.
2. Who validates the state that signatures accept as good. A high quorum on a single software stack provides no independence against a failure in that stack: this is the result Liquid left on September 6, 2026. The useful question is how many distinct implementations of the validator exist, not how many signatures are collected.
3. How long a patch takes to reach production. The release history of any project is public and can be read in a minute: publication dates, intervals between versions, and which files each one touches, as in the case of Elements.
4. What exactly the emergency plan covers. Models with timelocks and backup keys—Liquid uses a 4,032-block lock, about 28 days, for each peg-in UTXO, i.e., for each individual unspent balance—cover the disappearance of signers; an operation signed by procedure with a poorly validated state falls outside their scope.

The case also leaves a governance data point that does not depend on any secondary source: between September 6 and 10, 2026, the negotiation between the party controlling the funds and the issuer was conducted via OP_RETURN messages of 1,000 satoshis each. Blockstream's messages are signed in the clear with PGP and verified against the fingerprint 1176 542D A98E 71E1 3372 2EF7 4AC8 CC88 6844 A2D6; those from the other side are encrypted toward that same key. The crisis channel for a 326.20 million dollar network was the Bitcoin chain.

Sources and links: Blockstream API, L-BTC asset · Federation peg address · Liquid Federation incident report (Sep-8-2026, 19:10 UTC) · Defiprime: reconstruction of the peg-out and disputed block · SideSwap: reopening of markets and peg-ins · The Block: Blockstream refuses to pay reward · Blockstream: how the 11-of-15 multi-sig works · Elements PR 1600 (range proof cache) · Elements Releases · On-chain message log (automatically generated gist) · Bitquery: investigation of the 4,000 BTC peg-out on Liquid Network · Threshold: tBTC wallet generation · FBTC: custody · DefiLlama: protocol listing · Cryptotimes: Adam Back statements